Implementation of extreme learning machine algorithm in predicting the export value of crude palm oil to anticipate environmental damage in Bangka Belitung Islands Province.

Abstract The Bangka Belitung Islands Province is one of the crude palm oil (CPO) producing regions with a significant contribution to national exports. Sustainable policies that care about the environment in the field of exports and imports need to be directed to control the excessive expansion of oil palm land, by encouraging increased productivity of existing land and maintaining ecosystem sustainability. This study applied the Extreme Learning Machine (ELM) algorithm to predict the value of CPO exports in the region with time series data for the period January 2010 to December 2024. ELM was chosen for its ability to efficiently generate predictive models through random weighting and closed-form solutions using Moore-Penrose Inverses. This study used radial base activation function as well as three training-test data ratios (70:30, 80:20, 90:10) and showed that the radial base activation function with a 90:10 ratio of the 12-45-1 model gave the best results with RMSE of 0,115. The model obtained is then used to predict the value of exports over the next 12 months, showing a fluctuating trend with the lowest value predicted in June 2025 at $8.015.688 and the highest in April 2025 at $25.785.704. The results of this study show that the ELM method is effective in modelling and predicting the export value of CPO. It is hoped that this result can be the basis for the government to anticipate potential environmental damage through policies that balance economic growth and nature conservation.